The EU Digital Omnibus is dead, at least for the amendment cycle that compliance teams had been watching as a possible lifeline. The second political trilogue failed on April 28, and with it any realistic prospect of extending the August 2, 2026 deadline for Annex III high-risk AI systems. That deadline is established in primary EU law under Regulation (EU) 2024/1689, Article 85(3) and has not been modified. For providers and deployers of high-risk AI systems, particularly those in employment, critical infrastructure, and education – August 2 is the operative date.
According to IAPP reporting, the breakdown centered on disputes over Annex I sectoral rule mergers, including the Machinery and Toys Directives. The argument was structural: integrating those directives into the AI Act’s safety framework required reconciling sector-specific technical requirements that the three legislative bodies could not align in the available time. The specific sticking point, not the deadline itself, but the sectoral scope mechanics of Annex I, is what distinguishes this failure from the broader deadline-confirmation story that has been in circulation since late April.
Legal analysts have suggested that US Administration pressure on industry groups contributed to lobbying for the delay, though this characterization reflects legal inference rather than documented fact, per Holland & Knight’s analysis. What is documented is the result: negotiations failed, no extension was enacted, and the compliance calendar is unchanged.
The August 2 deadline now applies to the full scope of Annex III high-risk AI systems. That scope is broader than many deployers recognized a year ago. Per Epoch AI’s April 2026 compute tracking, 12 notable AI models already exceed the 10²⁵ FLOP threshold defined in Article 51 of the EU AI Act, the systemic risk designation that triggers GPAI provider obligations. Training compute for frontier models is doubling approximately every 5.2 months, and Epoch projects that 80 models will exceed 1×10²⁶ FLOP by 2028. The regulatory math has not kept pace with the deployment math.
For Annex III deployers, the operational checklist is unchanged but now has a firm endpoint. Risk management systems, technical documentation, human oversight mechanisms, and conformity assessment procedures must be in place before August 2. There is no indication that the EU AI Office is preparing discretionary enforcement grace periods, that is a characterization to watch for in official communications, but it has not been confirmed.
The collapse of the Omnibus negotiations does create one downstream consequence worth flagging: the GPAI Code of Practice process, which runs in parallel, is now the primary active mechanism for shaping how frontier model obligations are interpreted in practice. For organizations engaged in that process, the Omnibus failure may shift attention toward the Code rather than waiting for a legislative amendment that no longer has a near-term vehicle. Worth asking: if the negotiating breakdown is structural rather than political, what legislative format, if any, could realistically carry Annex I sectoral scope changes before August 2?